Lidar data in water resources applications

Transcription

1 Lidar data in water resources applications Paola Passalacqua CE 394K 3 Guest Lecture, November 15 th, 2012

2 Airborne Lidar Airborne laser altimetry technology (lidar, Light Detection And Ranging) provides highresolution topographical data, which can significantly contribute to a better representation of land surface. A valuable characteristic of this technology, which marks advantages over the traditional topographic survey techniques, is the capability to derive a high-resolution Digital Terrain Model (DTM) from the last pulse LiDAR data by filtering the vegetation points (Slatton et al., 2007). Slide courtesy of Dr. Paolo Tarolli, University of Padova, Italy

3 Airborne Lidar Slide courtesy of Dr. Paolo Tarolli, University of Padova, Italy

4 Airborne Lidar x,y,z Slide courtesy of Dr. Paolo Tarolli, University of Padova, Italy

5 Forest canopy Filter Processing First return/ Echo-waveform DSM 3D Buildings Raw LIDAR Multi-Spectral Filter Processing Land cover related to each vegetation types Filter Processing Last Return x,y,z DTM TIN Contour maps Slide courtesy of Dr. Paolo Tarolli, University of Padova, Italy

6 Topographic Lidar λ = 1064 nm Green LiDAR λ = 532 nm + λ =1064 nm Slide courtesy of Dr. Paolo Tarolli, University of Padova, Italy It is important to remember that the deep water surfaces normally do not reflect the signal: however this is not true in case of presence of floating sediments or when using bathymetric lidar. The bathymetric lidar, that is based on the same principles as topographic lidar, emits laser beams in two wavelengths: an infrared (1064 nm) and a green one (532 nm). The infrared wavelength is reflected on the water surface, while the green one penetrates the water and is reflected by the bottom surface or other objects in the water. Due to this reason the bathymetric lidar is also called green lidar.

7 Fonte: During optimal environment condition, when the water is clear, the green lidar survey may reach 50 m water depth with an horizontal accuracy of ±2.5 m, and vertical accuracy of ±0.25 m. This technology is growing fast, and some of the first applications in rivers are coming out (Hilldale and Raff, 2008; McKean et al., 2009). Slide courtesy of Dr. Paolo Tarolli, University of Padova, Italy

8 Example 1: Le Sueur River basin Minnesota River Basin Le Sueur River located in south-central Minnesota, covers an area of 2880 km 2 (87% row-crop agriculture) Provides ~ 24%-30% of the TSS entering the Minnesota River Le Sueur Lake Pepin Minnesota River major source of sediment for Lake Pepin (~85% of TSS load) Turbidity and related nutrients levels of Lake Pepin are far in excess of EPA standards State of Minnesota required to determine the sources of pollution and take management and policy actions NCED Research

9 Example 2: Limiting factors analysis of Coho salmon River networks produce a highly structured pattern of process and morphology downstream. This structure can be exploited to predict habitat and carrying capacity of species throughout the watershed. Ripple: spatially explicit model that links quantitatively topography, habitat carrying capacity and population dynamics for an entire watershed. GEO MODULE HAB MODULE POP MODULE GEO module: uses DEM to compute local slope and drainage area NCED Research

10 Example 3: Drainage density as a signature of climate Climate exerts a quantifiable control over the degree of channel dissection Lithology and relief are important cofactors Melton, M.A., ONR Tech Report, 1957 The problem of resolution: Strong discrepancy between channels as mapped in the field and as deduced from topographic maps (Morisawa, [1957; 1961]; Schneider [1961]). Schneider, W.J., J. Geophys. Res., 1961

11 Digital elevation data Data resolution available until recently m. Grigno basin, Italy Resolution 30 m x 30 m Data source: University of Padova Tirso basin, Italy Resolution 100 m x 100 m Data source: University of Padova Tanaro basin, Italy Resolution 90 m x 90 m Data source: University of Padova

12 Rio Cordon basin, Selva di Cadore, Italy Slide courtesy of Dr. Paolo Tarolli, University of Padova, Italy

13 The role of data resolution DTM 10x10 m Slide courtesy of Dr. Paolo Tarolli, University of Padova, Italy

14 The role of data resolution DTM 1x1 m Slide courtesy of Dr. Paolo Tarolli, University of Padova, Italy

15 Lidar DTMs: Solving the resolution problem? Availability of meter and sub-meter resolution topographic data Topographic patterns can be resolved over large areas at resolutions commensurate with the scale of governing processes Importance of objective extraction of geomorphic feature Opportunity to measure drainage density

16 Challenges in geomorphic feature extraction Channel initiation Identification of accurate centerline Presence of roads and bridges Artificial drainage ditches Small signal to noise ratio Identification of channel banks Measurement of bluffs

17 GeoNet: NCED toolbox for channel network extraction

18 GeoNet: Nonlinear filtering 1. Nonlinear filtering: Enhance features of interest, while smoothing small scale features. Perona and Malik [1990],,,, h x y t c x y t h t c 1 1 h / 2 Smooth Bumps Keep Channel

19 GeoNet: Statistical signature of geomorphic transitions 2. Skeleton of likely channelized pixels: Set of pixels with curvature above threshold, identified from quantile-quantile plot of curvature. k th The deviation of the pdf from Gaussian can be interpreted as transition from hillslope to valley [Lashermes et al,. 2007].

20 Channel extraction: geodesics 2. Geodesic minimization: Channels are extracted as paths of minimum cost The cost function ψ represents the cost of traveling between point a and point b in terms of a function of area (A), slope (S), curvature (κ) and skeleton (Skel): f 1 ( A, S,, Skel) 1 e. g., A b Channel head Outlet a

21 Channel initiation and channel disruptions Passalacqua, P., P. Tarolli, and E. Foufoula-Georgiou, Water Resour. Res, 2010

22 Flat lands and channel morphology Le Sueur River major source of sediment to the Minnesota River. Both listed as impaired for turbidity by USEPA. Need to identify sediment sources

23 Roads and ditches

24 Identification of likely channelized pixels in engineered landscapes Passalacqua, P., P. Belmont, and E. Foufoula-Georgiou, Water Resour. Res., 2012

25 Curvature analysis to distinguish channels and roads κ = γ = 2 h h h Passalacqua, P., P. Belmont, and E. Foufoula-Georgiou, Water Resour. Res., 2012

26 Differentiating natural versus artificial features Passalacqua, P., P. Belmont, and E. Foufoula-Georgiou, Water Resour. Res., 2012

27 Channel network extraction and bridge crossings Passalacqua, P., P. Belmont, and E. Foufoula-Georgiou, Water Resour. Res., 2012

28 Automatic extraction of channel morphology Height 60 m Automatic extraction of channel cross-section Detection of bank location Identification of geomorphic bankfull water surface elevation Measurements of channel width and of bank and bluff height

29 Automatic extraction of channel morphology Source: P. Belmont Source: C. Jennings Passalacqua, P., P. Belmont, and E. Foufoula-Georgiou, Water Resour. Res., 2012

30 Codependence of vegetation and drainage density Field mapped channel heads on slope gradient map (Imaizumi et al. [2010]). GeoNet drainage delineation.

31 Preliminary results: Codependence of vegetation and drainage density

32 Preliminary results: Codependence of vegetation and drainage density

33 Multi-resolution analysis of landscapes to understand landscape forming processes characteristic scales The first scaling break represents the length scale of the highly convex regions(i.e. the ridges) The second scaling break represents the hillslope length scale. Slide courtesy of Harish Sangireddy, UT Austin

34 Feature extraction from point cloud Six test sites were studied for testing the methodology. The data for the test sites was obtained from TNRIS Slide courtesy of Harish Sangireddy, UT Austin

35 Feature extraction from point cloud Site 1 has only streams and surrounding farmlands dry stream Slide courtesy of Harish Sangireddy, UT Austin

36 Feature extraction from point cloud Site 2 has roads, streams, marshy areas, small culvert and drains by the roadside stream marshy area road Slide courtesy of Harish Sangireddy, UT Austin

37 Feature extraction from point cloud Detects points around the water gaps and maps the geometry of stream properly Probable floodplain stream Slide courtesy of Harish Sangireddy, UT Austin

38 Feature extraction from point cloud When the elevation difference is very small in the region, the model identifies all low lying areas as water surfaces farmland drain road tree Site 5 Slide courtesy of Harish Sangireddy, UT Austin

39 Acknowledgements Work supported by: NSF EAR (NCED) NSF EAR NSF BCS / (PIs Stark and Passalacqua) Collaborators (in random order): Tien Do Trung, Colin Stark, Guillermo Sapiro, Bill Dietrich, Efi Foufoula-Georgiou, Nachiket Gokhale, Paolo Tarolli, Patrick Belmont, Harish Sangireddy, David Maidment, Yuichi S. Hayakawa, Fumitoshi Imaizumi, Tsuyoshi Hattanji, Lin Zhou.